kriti / benchmark.json
harrrshall's picture
release kriti nepali asr
762d1c1 verified
Raw
History Blame Contribute Delete
21.2 kB
{
"schema_version": 2,
"name": "kriti-nepali-asr-completed-snapshot-v1",
"scope": {
"evaluated_systems": 19,
"replicates_per_system": 2,
"records": 3630,
"view_sha256": "2374cac54831ce9c69282503763d7f1e12ada0404ae34ed471a7538cdae6c61f",
"sources": {
"fleurs_ne_np": 304,
"indicvoices_nepali": 2569,
"openslr54": 757
},
"primary_metric": "punctuation_insensitive_wer",
"tie_policy": "exact primary error ratios share a competition rank",
"selection_view": true
},
"systems": [
{
"rank": 1,
"key": "kriti",
"model": "kriti",
"model_id": "harrrshall/kriti",
"revision": "artifact-hashes-in-model.json",
"processor_id": null,
"processor_revision": null,
"backend": "ai4bharat_nemo_with_acoustic_danda_head",
"decoder": "rnnt",
"language": "ne",
"decode_configuration": {
"batch_fallbacks_allowed": false,
"batch_size": 32,
"compact_prediction_embedding": true,
"decoder": "rnnt",
"language_id": "ne",
"nepali_only_runtime_prune": true,
"punctuation_threshold": 0.711,
"sampling": false
},
"replicate_prediction_sha256": [
"1a42d4b0b527f2c21a4a28dfa84e7a2d769762bc4a6d80c59a12821e85b89f0f",
"1a42d4b0b527f2c21a4a28dfa84e7a2d769762bc4a6d80c59a12821e85b89f0f"
],
"punctuation_insensitive_wer": 0.2407729005728886,
"punctuation_insensitive_cer": 0.08287672638469905,
"raw_wer": 0.24685364976627114,
"raw_cer": 0.08404221560257552,
"artifact_sha256": {
"kriti.nemo": "0144854f0cc78f4b6115b75089fad632c39207d5256e53f92da996b9bbe43582",
"punctuation_head.json": "5874b6fc6b4f1172dffa249a42f5054ffe196cff9b97854fe180eafc4134e9bb"
},
"release_validation": {
"predictions_sha256": "1a42d4b0b527f2c21a4a28dfa84e7a2d769762bc4a6d80c59a12821e85b89f0f",
"benchmark_hypothesis_mismatches": 0
}
},
{
"rank": 1,
"key": "ai4bharat-ne-hybrid-rnnt",
"model": "AI4Bharat Nepali IndicConformer hybrid (RNNT)",
"model_id": "ai4bharat/indicconformer_stt_ne_hybrid_ctc_rnnt_large",
"revision": "cd09ba7720f3b17d259f6bfd03e1463bc5ba517d",
"processor_id": null,
"processor_revision": null,
"backend": "ai4bharat_nemo",
"decoder": "rnnt",
"language": "ne",
"decode_configuration": {
"batch_fallbacks_allowed": false,
"batch_size": 32,
"compact_prediction_embedding": false,
"decoder": "rnnt",
"language_id": "ne",
"nepali_only_runtime_prune": false,
"sampling": false
},
"replicate_prediction_sha256": [
"3656a93d28c3d1fb6dfea1c3ab7e3df6eb66a58820064e03ec331a2dee01f94b",
"3656a93d28c3d1fb6dfea1c3ab7e3df6eb66a58820064e03ec331a2dee01f94b"
],
"punctuation_insensitive_wer": 0.2407729005728886,
"punctuation_insensitive_cer": 0.08287672638469905,
"raw_wer": 0.25192776379399895,
"raw_cer": 0.08501127728471647
},
{
"rank": 3,
"key": "ai4bharat-ne-hybrid-ctc",
"model": "AI4Bharat Nepali IndicConformer hybrid (CTC)",
"model_id": "ai4bharat/indicconformer_stt_ne_hybrid_ctc_rnnt_large",
"revision": "cd09ba7720f3b17d259f6bfd03e1463bc5ba517d",
"processor_id": null,
"processor_revision": null,
"backend": "ai4bharat_nemo",
"decoder": "ctc",
"language": "ne",
"decode_configuration": {
"batch_fallbacks_allowed": false,
"batch_size": 32,
"compact_prediction_embedding": false,
"decoder": "ctc",
"language_id": "ne",
"nepali_only_runtime_prune": false,
"sampling": false
},
"replicate_prediction_sha256": [
"c47ad3a6bac4c1d7b46712cc9836d72dd71e90b2dbb472d65c1585fa92ba1340",
"c47ad3a6bac4c1d7b46712cc9836d72dd71e90b2dbb472d65c1585fa92ba1340"
],
"punctuation_insensitive_wer": 0.25310896860091425,
"punctuation_insensitive_cer": 0.08451503895695239,
"raw_wer": 0.26431339646010626,
"raw_cer": 0.08665595239313877
},
{
"rank": 4,
"key": "qwen3-asr-nepali-public",
"model": "Qwen3-ASR-Nepali (fine-tuned Nepali ASR)",
"model_id": "sidskarki/Qwen3-ASR-Nepali",
"revision": "6ae9dd5306dde33ff2507820c626bac6fb867a7f",
"processor_id": null,
"processor_revision": null,
"backend": "qwen3",
"decoder": null,
"language": "Nepali",
"decode_configuration": {
"batch_fallbacks_allowed": false,
"batch_size": 16,
"decoder": null,
"generation_prompt": null,
"language": "Nepali",
"max_new_tokens": null,
"precision": "backend_declared",
"sampling": false
},
"replicate_prediction_sha256": [
"f6a7268de4cc2bc2f3944453efcab325978e7b9551300731672308f8f0fa1ecc",
"f6a7268de4cc2bc2f3944453efcab325978e7b9551300731672308f8f0fa1ecc"
],
"punctuation_insensitive_wer": 0.5240433558895742,
"punctuation_insensitive_cer": 0.24217611689137886,
"raw_wer": 0.5551959726717008,
"raw_cer": 0.24943784185868761
},
{
"rank": 5,
"key": "whisper-large-v3-nepali-kiranpantha",
"model": "Whisper large-v3 Nepali (kiranpantha)",
"model_id": "kiranpantha/whisper-large-v3-nepali",
"revision": "934c2bfe2955282afec7c7ad9271b3e9ce6c7587",
"processor_id": "openai/whisper-large-v3",
"processor_revision": "06f233fe06e710322aca913c1bc4249a0d71fce1",
"backend": "whisper",
"decoder": null,
"language": "ne",
"decode_configuration": {
"batch_fallbacks_allowed": false,
"batch_size": 16,
"decoder": null,
"generation_prompt": null,
"language": "ne",
"max_new_tokens": 256,
"precision": "float32",
"sampling": false
},
"replicate_prediction_sha256": [
"a5210ed5016f1e46815c2132073a828f2c1f53d646212df48e51433d3579f0cb",
"a5210ed5016f1e46815c2132073a828f2c1f53d646212df48e51433d3579f0cb"
],
"punctuation_insensitive_wer": 0.5570593049484001,
"punctuation_insensitive_cer": 0.24941342253415147,
"raw_wer": 0.57836909185345,
"raw_cer": 0.253969228879403
},
{
"rank": 6,
"key": "whisper-large-v3-nepali-openslr",
"model": "Whisper large-v3 Nepali OpenSLR",
"model_id": "Dragneel/whisper-large-v3-nepali-openslr",
"revision": "e621bbb7a0ade7e047e2255cb947d2e4110d2466",
"processor_id": null,
"processor_revision": null,
"backend": "whisper",
"decoder": null,
"language": "ne",
"decode_configuration": {
"batch_fallbacks_allowed": false,
"batch_size": 16,
"decoder": null,
"generation_prompt": null,
"language": "ne",
"max_new_tokens": 256,
"precision": "float32",
"sampling": false
},
"replicate_prediction_sha256": [
"d0edc9cc89282843634bd124f86e220e9a37eaafda54047b279122a94a417e5a",
"d0edc9cc89282843634bd124f86e220e9a37eaafda54047b279122a94a417e5a"
],
"punctuation_insensitive_wer": 0.557678104477314,
"punctuation_insensitive_cer": 0.24670456333326493,
"raw_wer": 0.5745135642654521,
"raw_cer": 0.2506082226931043
},
{
"rank": 7,
"key": "nepali-asr-whisper-medium-paudeletal",
"model": "Whisper Medium Nepali (Paudel et al.)",
"model_id": "sumanpaudel1997/nepali-asr-whisper-medium",
"revision": "885f18aa455eedc1df0b99823977819f26e63d75",
"processor_id": "openai/whisper-medium",
"processor_revision": "abdf7c39ab9d0397620ccaea8974cc764cd0953e",
"backend": "whisper",
"decoder": null,
"language": "ne",
"decode_configuration": {
"batch_fallbacks_allowed": false,
"batch_size": 16,
"decoder": null,
"generation_prompt": null,
"language": "ne",
"max_new_tokens": 256,
"precision": "float32",
"sampling": false
},
"replicate_prediction_sha256": [
"194b9832531b87132798b3308287e932390d1fa88f4798ef5382fd8a3a72f7be",
"194b9832531b87132798b3308287e932390d1fa88f4798ef5382fd8a3a72f7be"
],
"punctuation_insensitive_wer": 0.5840269876439707,
"punctuation_insensitive_cer": 0.27638914540964654,
"raw_wer": 0.6086339845778896,
"raw_cer": 0.282310870893004
},
{
"rank": 8,
"key": "mms-1b-all-npi",
"model": "Meta MMS-1B-All (Nepali adapter)",
"model_id": "facebook/mms-1b-all",
"revision": "3d33597edbdaaba14a8e858e2c8caa76e3cec0cd",
"processor_id": null,
"processor_revision": null,
"backend": "mms",
"decoder": null,
"language": "npi",
"decode_configuration": {
"batch_fallbacks_allowed": false,
"batch_size": 32,
"decoder": null,
"generation_prompt": null,
"language": "npi",
"max_new_tokens": null,
"precision": "float32",
"sampling": false
},
"replicate_prediction_sha256": [
"dcbefc9587c2fd7ec2163c8576e1f5df4c02be843ff865aec895e76ff3afa26c",
"dcbefc9587c2fd7ec2163c8576e1f5df4c02be843ff865aec895e76ff3afa26c"
],
"punctuation_insensitive_wer": 0.6012735293530551,
"punctuation_insensitive_cer": 0.22785541806042944,
"raw_wer": 0.6127492109153382,
"raw_cer": 0.23010779105119308
},
{
"rank": 9,
"key": "nepali-asr-mms-1b",
"model": "MMS-1B Nepali (Paudel et al.)",
"model_id": "sumanpaudel1997/nepali-asr-mms-1b",
"revision": "1757b0fdf60ded4d5fa3294b63eba48ad4e5a32f",
"processor_id": null,
"processor_revision": null,
"backend": "ctc",
"decoder": null,
"language": "ne",
"decode_configuration": {
"batch_fallbacks_allowed": false,
"batch_size": 32,
"decoder": null,
"generation_prompt": null,
"language": "ne",
"max_new_tokens": null,
"precision": "float32",
"sampling": false
},
"replicate_prediction_sha256": [
"ffc35c3a19103d277d1304a1260c07fc29395be452b7eae79693bab6bdbb443e",
"ffc35c3a19103d277d1304a1260c07fc29395be452b7eae79693bab6bdbb443e"
],
"punctuation_insensitive_wer": 0.6076611373934567,
"punctuation_insensitive_cer": 0.24783667494373646,
"raw_wer": 0.6323065244316592,
"raw_cer": 0.25355976619680826
},
{
"rank": 10,
"key": "wav2vec2-xlsr-300m-nepali-shniranjan",
"model": "XLS-R 300M Nepali (shniranjan)",
"model_id": "shniranjan/wav2vec2-large-xlsr-300m-nepali",
"revision": "8b2f97cbca949f7d6d805ff5e03b62b82d106198",
"processor_id": null,
"processor_revision": null,
"backend": "ctc",
"decoder": "greedy",
"language": "ne",
"decode_configuration": {
"batch_fallbacks_allowed": false,
"batch_size": 32,
"decoder": "greedy",
"generation_prompt": null,
"language": "ne",
"max_new_tokens": null,
"precision": "float32",
"sampling": false
},
"replicate_prediction_sha256": [
"e0fefd12ecd5093565b42743adbb962a7343597a6c586bd163038dcac74e38a1",
"e0fefd12ecd5093565b42743adbb962a7343597a6c586bd163038dcac74e38a1"
],
"punctuation_insensitive_wer": 0.6288999341277921,
"punctuation_insensitive_cer": 0.3343525758104346,
"raw_wer": 0.639038715090495,
"raw_cer": 0.33592659698976685
},
{
"rank": 11,
"key": "indicconformer-ne-hybrid",
"model": "IndicConformer Nepali hybrid",
"model_id": "sumanpaudel1997/nepali-asr-conformer-hi",
"revision": "c3bfe5f21003224045a65705568a4cdbc906f71e",
"processor_id": null,
"processor_revision": null,
"backend": "nemo",
"decoder": null,
"language": "ne",
"decode_configuration": {
"batch_fallbacks_allowed": false,
"batch_size": 32,
"decoder": null,
"language": "ne",
"sampling": false
},
"replicate_prediction_sha256": [
"7fc6129c5c85ec7f2f2c2c74d7e46368c26024f0483d1d661757307699b14814",
"7fc6129c5c85ec7f2f2c2c74d7e46368c26024f0483d1d661757307699b14814"
],
"punctuation_insensitive_wer": 0.7036549094756173,
"punctuation_insensitive_cer": 0.34641589197397854,
"raw_wer": 0.7326501258540093,
"raw_cer": 0.35395660378002297
},
{
"rank": 12,
"key": "seamless-m4t-v2-large-npi",
"model": "Meta SeamlessM4T v2 large",
"model_id": "facebook/seamless-m4t-v2-large",
"revision": "5f8cc790b19fc3f67a61c105133b20b34e3dcb76",
"processor_id": null,
"processor_revision": null,
"backend": "seamless",
"decoder": null,
"language": "npi",
"decode_configuration": {
"batch_fallbacks_allowed": false,
"batch_size": 8,
"decoder": null,
"generation_prompt": null,
"language": "npi",
"max_new_tokens": null,
"precision": "backend_declared",
"sampling": false
},
"replicate_prediction_sha256": [
"eb764a7349273e81e9a3da6957fe6e96dabe35c23d3af4051d8a3066361d1eab",
"eb764a7349273e81e9a3da6957fe6e96dabe35c23d3af4051d8a3066361d1eab"
],
"punctuation_insensitive_wer": 0.7042337864542787,
"punctuation_insensitive_cer": 0.4389754218911394,
"raw_wer": 0.7140716768548484,
"raw_cer": 0.44084799721565376
},
{
"rank": 13,
"key": "nepali-asr-xlsr-53-paudeletal",
"model": "XLS-R 53 Nepali (Paudel et al.)",
"model_id": "sumanpaudel1997/nepali-asr-xlsr-53",
"revision": "36151186fa385cd83c23969a68e847b90ca377fd",
"processor_id": null,
"processor_revision": null,
"backend": "ctc",
"decoder": null,
"language": "ne",
"decode_configuration": {
"batch_fallbacks_allowed": false,
"batch_size": 32,
"decoder": null,
"generation_prompt": null,
"language": "ne",
"max_new_tokens": null,
"precision": "float32",
"sampling": false
},
"replicate_prediction_sha256": [
"759c15b75e93190ebb7c1d8f74bcfdaa71f88987b94cef6b7798b0722051cedd",
"759c15b75e93190ebb7c1d8f74bcfdaa71f88987b94cef6b7798b0722051cedd"
],
"punctuation_insensitive_wer": 0.7499451064934028,
"punctuation_insensitive_cer": 0.2936410214314542,
"raw_wer": 0.7648727476127691,
"raw_cer": 0.2976520727342212
},
{
"rank": 14,
"key": "wav2vec2-xlsr-300m-nepali-spktsagar",
"model": "XLS-R 300M Nepali OpenSLR (spktsagar)",
"model_id": "spktsagar/wav2vec2-large-xls-r-300m-nepali-openslr",
"revision": "bcda4d6038be2fdbd8b7af0f6dc7be89e60506a5",
"processor_id": null,
"processor_revision": null,
"backend": "ctc",
"decoder": null,
"language": "ne",
"decode_configuration": {
"batch_fallbacks_allowed": false,
"batch_size": 32,
"decoder": null,
"generation_prompt": null,
"language": "ne",
"max_new_tokens": null,
"precision": "float32",
"sampling": false
},
"replicate_prediction_sha256": [
"15adcb1289f57fa6efeec890f2c2433f3893350a1861119d981dc813637c3f90",
"15adcb1289f57fa6efeec890f2c2433f3893350a1861119d981dc813637c3f90"
],
"punctuation_insensitive_wer": 0.882687586082999,
"punctuation_insensitive_cer": 0.43844185871520724,
"raw_wer": 0.8864117623556674,
"raw_cer": 0.43971856264949655
},
{
"rank": 15,
"key": "whisper-large-v3",
"model": "Whisper large-v3",
"model_id": "openai/whisper-large-v3",
"revision": "06f233fe06e710322aca913c1bc4249a0d71fce1",
"processor_id": null,
"processor_revision": null,
"backend": "whisper",
"decoder": null,
"language": "ne",
"decode_configuration": {
"batch_fallbacks_allowed": false,
"batch_size": 16,
"decoder": null,
"generation_prompt": null,
"language": "ne",
"max_new_tokens": 256,
"precision": "float32",
"sampling": false
},
"replicate_prediction_sha256": [
"ea16f58b9106b2912e2f6559c002248a825497869dd84be2bd9a2d2af786e1e0",
"ea16f58b9106b2912e2f6559c002248a825497869dd84be2bd9a2d2af786e1e0"
],
"punctuation_insensitive_wer": 0.9702177775116274,
"punctuation_insensitive_cer": 0.43327724079432506,
"raw_wer": 0.9755483638978785,
"raw_cer": 0.4411755673617296
},
{
"rank": 16,
"key": "vakyansh-wav2vec2-nepali-nem-130",
"model": "Vakyansh wav2vec2 Nepali",
"model_id": "Harveenchadha/vakyansh-wav2vec2-nepali-nem-130",
"revision": "002c6200bcdacd8558e696583361f47e8154df67",
"processor_id": null,
"processor_revision": null,
"backend": "ctc",
"decoder": null,
"language": "ne",
"decode_configuration": {
"batch_fallbacks_allowed": false,
"batch_size": 32,
"decoder": null,
"generation_prompt": null,
"language": "ne",
"max_new_tokens": null,
"precision": "float32",
"sampling": false
},
"replicate_prediction_sha256": [
"5c69336150d7c1d68d61e8b42909a9b4cac5683eaf92719aced10beab2953a04",
"5c69336150d7c1d68d61e8b42909a9b4cac5683eaf92719aced10beab2953a04"
],
"punctuation_insensitive_wer": 1.0157494460746153,
"punctuation_insensitive_cer": 1.987769090274785,
"raw_wer": 1.0160214151584162,
"raw_cer": 1.9832393275257876
},
{
"rank": 17,
"key": "whisper-large-v3-turbo",
"model": "Whisper large-v3-turbo",
"model_id": "openai/whisper-large-v3-turbo",
"revision": "41f01f3fe87f28c78e2fbf8b568835947dd65ed9",
"processor_id": null,
"processor_revision": null,
"backend": "whisper",
"decoder": null,
"language": "ne",
"decode_configuration": {
"batch_fallbacks_allowed": false,
"batch_size": 16,
"decoder": null,
"generation_prompt": null,
"language": "ne",
"max_new_tokens": 256,
"precision": "float32",
"sampling": false
},
"replicate_prediction_sha256": [
"89f4ef288b70c14650fce36a099d6eeb2fdeaa3d5190b0408ed8a9b906e5e365",
"89f4ef288b70c14650fce36a099d6eeb2fdeaa3d5190b0408ed8a9b906e5e365"
],
"punctuation_insensitive_wer": 1.1123620176856897,
"punctuation_insensitive_cer": 0.5548783407553339,
"raw_wer": 1.1184026529226099,
"raw_cer": 0.5606977244111415
},
{
"rank": 18,
"key": "qwen3-asr-0.6b-base",
"model": "Qwen3-ASR 0.6B base",
"model_id": "Qwen/Qwen3-ASR-0.6B",
"revision": "5eb144179a02acc5e5ba31e748d22b0cf3e303b0",
"processor_id": null,
"processor_revision": null,
"backend": "qwen3",
"decoder": null,
"language": "Nepali",
"decode_configuration": {
"batch_fallbacks_allowed": false,
"batch_size": 16,
"decoder": null,
"generation_prompt": null,
"language": "Nepali",
"max_new_tokens": null,
"precision": "backend_declared",
"sampling": false
},
"replicate_prediction_sha256": [
"88e8bdeaf83d2f62a2ed27fe7f53b5221bfa45882331d95e4c25db82b522ff02",
"88e8bdeaf83d2f62a2ed27fe7f53b5221bfa45882331d95e4c25db82b522ff02"
],
"punctuation_insensitive_wer": 1.120206798810308,
"punctuation_insensitive_cer": 0.7182752228310315,
"raw_wer": 1.1213592233009708,
"raw_cer": 0.7254450347531451
},
{
"rank": 19,
"key": "whisper-large-v3-turbo-nepali-kiranpantha",
"model": "Whisper large-v3-turbo Nepali (kiranpantha)",
"model_id": "kiranpantha/whisper-large-v3-turbo-nepali",
"revision": "9e8f4b238b696a14f9c20c8e01b8f3ef36172399",
"processor_id": "openai/whisper-large-v3-turbo",
"processor_revision": "41f01f3fe87f28c78e2fbf8b568835947dd65ed9",
"backend": "whisper",
"decoder": null,
"language": "ne",
"decode_configuration": {
"batch_fallbacks_allowed": false,
"batch_size": 16,
"decoder": null,
"generation_prompt": null,
"language": "ne",
"max_new_tokens": 256,
"precision": "float32",
"sampling": false
},
"replicate_prediction_sha256": [
"ffc46aa64b592dbc565e04a87c9b5efab9178911ac20df1f4b652dfa2c183ad8",
"ffc46aa64b592dbc565e04a87c9b5efab9178911ac20df1f4b652dfa2c183ad8"
],
"punctuation_insensitive_wer": 1.372437471305667,
"punctuation_insensitive_cer": 4.323889949174688,
"raw_wer": 1.3721882616165249,
"raw_cer": 4.3190123760095815
}
],
"limitations": "Development-set selection evidence; not an untouched test or universal domain claim."
}